PERSON DETECTION IN DAYTIME AND NIGHTTIME ENVIRONMENTS FOR SEARCH AND PATROL MISSIONS
PERSON DETECTION IN DAYTIME AND NIGHTTIME ENVIRONMENTS FOR SEARCH AND PATROL MISSIONS

Author(s): Pedro MARGALHO, Gonçalo CRUZ
Subject(s): Politics / Political Sciences, Politics, Security and defense
Published by: Carol I National Defence University Publishing House
Keywords: Human Detection; Patrol; Search and Rescue; Unmanned Aircraft Systems;
Summary/Abstract: The search and rescue mission in Portugal is carried out by several entities, including the Portuguese Air Force (PAF). This mission can be significantly prolonged due to various factors, leading to higher-than-expected costs and, in extreme cases, causing fatigue among pilots and crew members. Furthermore, the FA is also concerned with patrolling areas of interest, such as airbases and squadrons. Considering this, the study's objectives are to develop an automatic person detection system capable of being used in both daytime and nighttime environments for real-time automated analysis and decision support in search and patrolling missions using low-altitude aerial images. The system was based on convolutional neural networks (CNN) using YOLOv8m. Several models were trained using publicly available datasets and additional data collection conducted to complement the military context. The models were subsequently validated on specific test sets, representing each type of scenario in line with the study’s objectives.

• Details
• Contents